Relationship of retinal arteriolar (central retinal arteriolar equivalent [CRAE]) and venular (central retinal venular equivalent [CRVE]) calibers to current mean arterial blood pressure (MABP), smoking, body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) and serum albumin level. Vertical bars indicate 95% confidence intervals. CRAE and CRVE are adjusted for the factors in model 2.
Liew G, Sharrett AR, Wang JJ, Klein R, Klein BEK, Mitchell P, Wong TY. Relative Importance of Systemic Determinants of Retinal Arteriolar and Venular CaliberThe Atherosclerosis Risk in Communities Study. Arch Ophthalmol. 2008;126(10):1404-1410. doi:10.1001/archopht.126.10.1404
To examine the relative contributions of systemic cardiovascular factors to retinal arteriolar and venular caliber in men and women and in whites and African Americans.
In the Atherosclerosis Risk in Communities study, retinal arteriolar caliber (central retinal arteriolar equivalent), and venular caliber (central retinal venular equivalent) were measured from digitized retinal photographs of 8794 participants.
The main systemic determinants of narrower central retinal arteriolar equivalent were, in order of relative decreasing contribution, higher current mean arterial blood pressure, lower serum albumin level, current alcohol consumption, and higher body mass index (calculated as weight in kilograms divided by height in meters squared). The main systemic determinants of wider central retinal venular equivalent were current cigarette smoking and higher current mean arterial blood pressure, followed by higher white blood cell count, body mass index, and plasma low-density lipoprotein cholesterol levels. These associations were generally similar in whites and African Americans and in men and women.
The major systemic determinant of narrower retinal arteriolar caliber is higher blood pressure, while those of wider retinal venular caliber are cigarette smoking, higher blood pressure, systemic inflammation, and obesity. These data offer further insights into the systemic processes influencing arteriolar and venular characteristics and may help explain the observed associations of retinal vascular caliber and the risk of clinical cardiovascular disease.
The retina is a unique site where the small blood vessels of the human vasculature can be directly and noninvasively visualized. As such, it offers an opportunity for detailed in vivo study of the structure of small arterioles and venules and their association with systemic disease processes such as hypertension and diabetes.
Recent studies have investigated the relationship between systemic disease factors and retinal arteriolar and venular caliber.1- 9 However, several important issues remain unclear. First, few studies have examined the relative importance of risk factors in determining arteriolar and venular calibers.10 This information may improve understanding of microvascular disease pathology and suggest the most appropriate approach to therapeutic options targeted at microcirculation. Second, whether the risk factors of vessel caliber vary by racial and/or ethnic groups and by sex is unclear.4- 6,11 Finally, it is unknown whether past exposures to major systemic factors (eg, hypertension, hyperglycemia, cigarette smoking) influence retinal vessel caliber.
The current analysis of data from the Atherosclerosis Risk in Communities (ARIC) Study was designed to address all of these issues. We examined the relative importance of both current and past (eg, past blood pressure, past glucose levels, past smoking) systemic determinants of retinal arteriolar and venular caliber. We also examined whether associations were similar between whites and African Americans and between men and women.
The ARIC study is a population-based cohort study that included 15 792 women and men aged 45 to 64 years at recruitment from 1987 through 1989.12,13 The study sample is composed of individuals who participated in the third examination, when retinal photography was first performed. Of the 12 887 who participated in this examination, we excluded 211 persons who were not present at the second examination, 38 whose race was neither African American nor white, 42 residents of color in Minneapolis and Washington County, Minnesota (owing to small numbers), 224 with no retinal photographs, 1774 with ungradeable photographs, 21 with retinal vein occlusions, 981 without data on carotid and/or intima media thickness, and 802 without data on serum albumin level, leaving 8794 with data available for the current analysis. Persons excluded were slightly older (by 1 year), more likely to be African American, and had higher mean arterial blood pressure (MABP) at visit 3 (by 2 mm Hg), but did not differ by sex from persons included. Institutional review boards at each study site and at the Fundus Photograph Reading Center at the University of Wisconsin–Madison approved the study. Informed consent was obtained from all participants and the study was conducted in accordance with the Declaration of Helsinki. Differences between participants and nonparticipants at the baseline examination are presented elsewhere.14
The retinal photography procedure and grading of retinal microvascular signs are described in detail elsewhere.15 Briefly, a 45-degree retinal photograph of 1 randomly selected eye of each participant was taken at visit 3 following 5 minutes of dark adaptation. This photograph was centered on the region of the optic disc and the macula and was taken using an autofocus camera.
Trained graders masked to participant characteristics used a computer-assisted approach to measure the calibers of all arterioles and venules coursing through a specified area surrounding the optic disc.15 Individual vessel measurements were combined into summary indices—the central retinal arteriole equivalent (CRAE) and the central retinal venule equivalent (CRVE). These indices represent the estimated central retinal arteriolar and venular caliber of the eye after taking into account the branching patterns.15 These measurements are reliable, with intragrader and intergrader reliability coefficients for CRAE and CRVE of 0.84 and 0.79, respectively.15,16 Intraindividual reliability coefficients were also generally high.16
At each examination, blood pressures were taken with a random-zero sphygmomanometer and MABP was calculated as two-thirds of the diastolic plus one-third of the systolic value.12,13,17 Current MABP was defined as the MABP value from visit 3, while past MABP was defined as the average MABP values from visits 1 and 2. Cigarette smoking and alcohol consumption were ascertained from a questionnaire. Mean carotid and/or intima media thickness was derived from standardized B-mode ultrasonograms.18 Total plasma cholesterol and triglyceride levels were measured by enzymatic methods.19 Measurement of other variables is described elsewhere.17,20
To assess the determinants of CRAE and CRVE, we used multiple linear regression models that adjusted for continuous variables (per interquartile range [IQR]) and dichotomous variables (presence vs absence) and compared the partial R2, a measure of the contribution of each systemic factor to the total variance in CRAE or CRVE. We used the IQR to provide a standardized comparison between continuous variables, as some were normally distributed while others were not (eg, serum glucose). We constructed 3 multivariable models. Model 1 adjusted for the variables in Table 1. Narrower arteriolar caliber and wider venular caliber may be due to pathology, normal anatomical variation, or measurement error, and we have previously shown that anatomical variation and measurement error in arteriolar and venular caliber may lead to spurious associations.21,22 We attempted to control for these factors by including the fellow vessel caliber in the models as a covariable, as both arteriolar and venular calibers are influenced by these factors in the same manner (ie, persons of smaller stature have narrower arterioles and venules, and measurement errors from ocular magnification and background pigmentation tend to overestimate or underestimate both arteriolar and venular calibers in a similar manner). Model 2 thus adjusted for variables in model 1 as well as the fellow vessel caliber (ie, addition of CRAE in models for CRVE, and CRVE in models for CRAE). Model 3 adjusted for all of the variables in model 2 as well as past MABP, past serum glucose level, past plasma triglyceride level, past smoking, and past alcohol consumption. We report differences in CRAE and CRVE (μm), 95% confidence intervals, and partial R2. We repeated these analyses stratified by race and sex. For all analyses, SAS version 8.2 (SAS Institute Inc, Cary, North Carolina) was used.
The distribution of vascular risk factors in the population is described in Table 1. African Americans had wider CRAE (approximately 3 μm) and CRVE (approximately 8-9 μm) than whites. Women had slightly wider CRAE (2 μm) and narrower CRVE (2 μm) than men; partici pants aged 60 years or older had slightly narrower (2-3 μm) CRAE and CRVE (data not shown).
We examined the relationships of CRAE and CRVE with systemic factors according to model 1 (unadjusted for fellow vessel caliber) and model 2 (adjusted for fellow vessel caliber). In general, both models gave similar results, but there were a few exceptions. Associations for CRVE with increasing MABP reversed in direction after additional adjustment in model 2, while associations for CRAE with serum glucose, white blood cell count, and current smoking were reduced in magnitude and lost statistical significance. For example, with each IQR increase in MABP (ie, comparing a person with higher current MABP [eg, in the 75th percentile] with a person with low current MABP [eg, in the 25th percentile]), CRAE decreased by 7.0 μm in model 1. This is interpreted to mean that a person with higher current MABP in the 75th percentile had a CRAE that was, on average, 7.0 μm narrower (4.0% narrower) than that of a person with lower current MABP (25th percentile); a similar interpretation should be applied for results in this and following tables. The MABP-CRAE association changed little with additional adjustment for CRVE. With each IQR increase in MABP, CRVE decreased by 1.0 μm in model 1. However, after additional adjustment for CRAE in model 2, the association per IQR increase in current MABP reversed in direction from −1.0 to 2.6 μm. The results after adjusting for model 2 are presented in Table 2.
The greatest proportional determinants of narrower CRAE, as measured from partial R2 contribution to model 2, were higher current MABP (0.089), lower serum albumin level (0.003), and current alcohol consumption (0.002) (Table 3). The largest determinants of wider CRVE by partial R2 were current smoking status (0.026), higher current MABP (0.026), higher white blood cell count (0.006), body mass index (calculated as weight in kilograms divided by height in meters squared) (0.002), and low-density lipoprotein cholesterol level (0.002). Persons with higher values of current MABP (75th percentile) had CRVE that, on average, was 2.6 μm wider (1.4% wider) than persons with lower values of current MABP (25th percentile).
Systemic associations of CRAE were similar in whites and African Americans. Most associations with CRAE in men and women were similar. For example, men with higher current MABP (75th percentile) had CRAE that was 6.5 μm narrower (4.1% narrower) than men with lower current MABP (25th percentile), while women with higher current MABP had CRAE that was 6.3 μm narrower than women with low current MABP (4.0%). There were some sex differences, particularly with high-density lipoprotein cholesterol level, endothelial function markers, and current alcohol consumption (data not shown).
Persons with higher current MABP (75th percentile) had CRAE that was 4.8 μm narrower (3.0% narrower) than persons with lower MABP (25th percentile), whereas persons with higher past MABP (75th percentile) had CRAE that was only 2.6 μm narrower (1.6% narrower) than persons with lower past MABP (25th percentile) (Table 4). Higher current and past MABP were both associated with wider CRVE, as were current and past smoking.
In a supplementary analysis, we excluded the 2297 participants taking antihypertensive treatments. The association of narrower CRAE with current MABP strengthened slightly, with CRAE narrowing by 7.4 μm per IQR increase in MABP in model 1.
In summary, most associations of CRAE and CRVE were similar using model 1 or model 2. The major proportional determinants of narrower CRAE were higher current MABP, lower serum albumin level, and current alcohol consumption (Figure). The major determinants of wider CRVE were current smoking, higher current MABP, higher white blood cell count, body mass index, and low-density lipoprotein cholesterol level (Figure). Associations were largely similar across race and sex.
The ARIC study has well-documented data on a range of systemic cardiovascular risk factors and digitized retinal images, from which we measured retinal arteriolar (CRAE) and venular (CRVE) caliber. In an earlier study, our group examined the association of cardiovascular risk factors with arteriole to venule ratio, a measure of arteriolar narrowing that has since been shown to have several limitations.21 Our findings are consistent with those from our earlier study and provide further insight into the vascular processes driving changes in the arteriole to venule ratio.
Our study expands on earlier work examining the specific relationships of cardiovascular, metabolic, inflammatory, and lifestyle risk factors with retinal vascular caliber,17 including reports from the Multi-Ethnic Study of Atherosclerosis (MESA),3 Beaver Dam Eye Study (BDES),4 Rotterdam Study,5 Blue Mountains Eye Study,6,11 Cardiovascular Health Study,7 and the Funagata Study.8 All studies consistently found that high blood pressure was associated with narrower arteriolar caliber and that smoking and body mass index were associated with wider venular caliber. Some studies have also reported high blood pressure associated with narrower venular caliber and smoking associated with wider arteriolar caliber, which, our analyses now show, disappear with adjustment for natural anatomic variation and measurement error.5 In the case of both MABP and smoking, findings from models with additional adjustment may be more congruent with clinical experience and existing biological knowledge. For example, wider retinal venules with increasing MABP may be explained as a response to chronic retinal hypoxia or hypoperfusion23,24 from chronic hypertensive damage to the microcirculation, whereas it is less biologically apparent why narrower venules should be associated with increasing MABP. Similarly, a lack of influence of smoking on arteriolar caliber is more consistent with the literature that states that smoking has little effect on arteriolar caliber or causes only mild arteriolar narrowing25,26 rather than widening.
An important concern in controlling for anatomy and measurement error using the fellow vessel caliber is the potential for overadjustment where anatomic factors com mon to both arteriolar and venular caliber may be inappropriately adjusted for. This is particularly relevant in determining the associations of age, sex, and race on vascular caliber. Increasing age was a major contributor to narrower arteriolar and venular caliber. Other researchers have reported that men have narrower arterioles than women,3,27 which could be related to the higher blood pressure in men (with higher arteriolar resistance).27 In animal experiments, male sex is associated with narrower arterioles than female sex at the same blood pressure level.28,29 The associations of male sex and African American race with wider venular caliber were reported in the BDES4 and MESA,3 respectively, but the underlying mechanisms are not known. These differences may reflect genetic or cardiovascular influences not examined in this study.
We observed that the main determinants of narrower arteriolar caliber (other than demographic factors) were higher current MABP, lower serum albumin level, and current alcohol consumption. In longitudinal analyses, both current and past MABP were strongly associated with narrower arteriolar caliber. All studies to date have consistently demonstrated that current3- 8 and past blood pressure30,31 are strongly associated with narrower arte riolar caliber, confirming retinal arteriolar narrowing as a marker of current and longer-term blood pressure levels. The association of lower serum albumin level with narrower retinal arterioles was also observed in the BDES,4 although it was not statistically significant in that population. Lower serum albumin level may be a nonspecific marker of inflammation or chronic kidney disease, and the association with narrower arterioles may reflect these disease processes. The markers of endothelial dysfunction examined in this study, von Willebrand factor and factor VIII levels, made at most small or negligible contributions to arteriolar caliber, consistent with results from the BDES,4 MESA,3 and Hoorn studies32 that found no association of retinal arteriolar caliber with other, more specific serum markers of endothelial dysfunction (eg, serum soluble intercellular adhesion molecule–1). However, we did not have data on other markers of endothelial dysfunction, such as plasma asymmetric dimethylarginine.33 The association with current alcohol consumption has been reported in some studies (MESA),3 but not others (Rotterdam),5 and may reflect impaired vasodilation with chronic alcohol ingestion, which is associated with impaired vasodilation in the cerebral34,35 and systemic microcirculations.36
The main determinants of wider venular caliber in this population (other than demographic factors) were current smoking and higher current MABP, followed by increasing white blood cell count. Smoking has consistently been associated with wider retinal venules in the Rotterdam study,5 BDES,4 MESA,3 Wisconsin Epidemiologic Study of Diabetic Retinopathy,10 and Blue Mountains Eye Study,37 and we now additionally show a temporal association of past smoking with wider retinal venules that is independent of current smoking status. Retinal venule widening in smokers has been observed clinically, and may be due to reduced oxyhemoglobin and tissue hypoxia, nicotine-induced changes in vessel autoregulation, and secondary polycythaemia.38,39 Chronic smoking also promotes inflammation and endothelial dysfunction,40- 42 which may lead to disruptions of microvascular autoregulatory vasomotor function and venular widening.43,44 Wider retinal venules may thus be a lasting microvascular marker of damage from long-standing smoking.
Higher white blood cell count was a major contributor to wider venular caliber, with lesser contributions from other nonspecific markers of inflammation (fibrinogen and serum albumin levels).3- 5 A range of inflammatory markers has now been consistently linked with wider retinal venules in all the major population-based studies,3- 5,45 strongly implicating a role for inflammation in influencing retinal venular caliber. Inflammation may cause venular widening through mechanisms such as damage to the venular endothelial surface layer,5,45 resulting in loss of glycocalyx (and hence a wider blood column detected on retinal photography) or increased vasodilation.
We also examined for possible racial and/or ethnic and sex differences in associations. Associations were largely similar, but some systemic determinants appeared to vary by race or sex, eg, body mass index, alcohol consumption, and markers of endothelial dysfunction. These findings should be interpreted with caution, because they may be chance associations, they were not specified a priori, most interactions were not strongly statistically significant, and the mechanisms that may underlie them are not known.
The vascular risk factors examined in this study explain more of the variation in retinal arteriolar caliber than in retinal venular caliber, highlighting the lack of knowledge about the determinants of venular caliber. It was assumed earlier that venular calibers remain relatively static in the presence of vascular disease, but this is now shown to clearly not be the case. Some recent studies indicate that venular widening may be a marker of other processes, such as endothelial dysfunction,3 hypoperfusion, and cerebral hypoxia.46 However, it should be borne in mind that associations with venular widening may actually reflect associations with arteriolar widening; owing to the strong concomitant association of arteriolar narrowing with blood pressure, this possibility is difficult to determine, and our method of adjusting for the fellow vessel caliber may be an overadjustment.
Strengths of this study include the use of well-measured cardiovascular risk factors, adjustment for many potential confounders, and large sample size permitting subgroup analyses. As the main analyses were cross-sectional, care must be taken in interpreting the temporal sequence of effect. Although many associations were observed, most contributed only marginally to differences in vascular caliber. Although a high proportion of subjects had ungradable photographs, we do not believe this influenced the associations we observed, as it is unlikely that persons with ungradeable photographs had retinal vessel calibers different from persons with gradable photographs.
In summary, findings from the ARIC population suggest distinct patterns of systemic associations with arteriolar and venular caliber. The major systemic determinants of narrower arteriolar caliber were higher current blood pressure, lower serum albumin, and current alcohol consumption. The major determinants of wider venular caliber were current smoking and higher blood pressure, with contributions from inflammatory processes. Past blood pressure and past smoking were associated with narrower arteriolar caliber and wider venular caliber, respectively. Our results also indicate that spurious associations from shared anatomy and measurement error may occur regarding associations with blood pressure and smoking. However, regarding other systemic variables, spurious associations do not appear to be a major issue. These data show clearly the deleterious effects of elevated blood pressure, smoking, alcohol, obesity, dyslipidemia, and inflammation on microcirculation and emphasize again the importance of controlling these risk factors to prevent and treat cardiovascular disease. More research is needed into whether close monitoring of the retinal microvasculature using retinal imaging techniques permits clinicians to quantify the efficacy of treatments of elevated blood pressure and other vascular risk factors.
Correspondence: Tien Yin Wong, MD, PhD, Centre for Eye Research Australia, University of Melbourne, 32 Gisborne St, East Melbourne 3002, Australia (firstname.lastname@example.org).
Submitted for Publication: November 15, 2007; final revision received June 19, 2008; accepted June 25, 2008.
Author Contributions: Tien Wong had full access to all of the data in this study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Financial Disclosure: None reported.
Funding/Support: This study was supported by contracts N01-HC-35125, N01-HC-35126, N01-HC-55015, N01-HC-55016, N01-HC-55018, N01-HC-55019, N01-HC-55020, N01-HC-55021, and N01-HC-55022 from the National Heart, Lung, and Blood Institute, Bethesda, Maryland. Additional support was provided by the National Heart Foundation, the Science Technology Innovation Grant, Victoria, and the Sylvia and Charles Viertel Clinical Investigator Award (Dr Wong).
Additional Information: The authors thank the staff and participants of the Atherosclerosis Risk in Communities study for their important contributions.